TECHNICAL FIELD
The present invention relates to a picture signal coding apparatus for
reducing a data quantity of a picture signal and coding it, aiming to reduce a
capacity necessary for recording and a transmission rate necessary for
transmission at recording or transmission of the picture signal and a picture
signal decoding apparatus for correctly decoding the coded picture signal.
TECHNICAL BACKGROUND
As a coding method used in a picture signal coding apparatus for a
natural picture, it is well known that JPEG (Joint Photographic coding
Experts Group) and MPEG (Moving Picture Experts Group) are good method
in high effective coding. In each method, an input picture signal is divided
into rectangular blocks made of a plurality of pixels, then the picture signal
at every block is orthogonally transformed (discrete cosine transform) by
orthogonal transform means, and after being quantized through specified
quantization steps at quantization means, the picture signal is variable
length coded at variable length coding means and is output as a coded
picture signal.
On the other hand, as a picture signal, there is a synthesized picture
signal made by artificially synthesizing a plurality of pictures other than a
usual picture signal composing a picture. In such a synthesized picture, if a
picture signal before synthesis is coded at a picture coding apparatus, the
picture signal before synthesis and the picture signal after synthesis are
arbitrarily selected and are decoded at a picture decoding apparatus and a
picture can be reproduced and is available for a picture data base and the
like. For a synthesized picture, a signal called transparency information
(shape information) expressing a ratio of picture synthesis is necessary other
than a color signal composed of a luminance signal and color difference
signals. If the transparency information is 100 %, the picture signal implies
to be transparent and there is no need of coding. Therefore, if a pixel value is
coded only for an opaque picture element which the transparency
information is not 100 %, an efficient coding can be done. Because a gather of
the opaque pixels is generally not rectangular shape, it is impossible to use
discrete cosine transform used in JPEG and MPEG as it is. As for a practical
method of shape adaptive orthogonal transformation which can be efficiently
coded even in any shape other than rectangle, investigation results are
reported in Coding of Arbitrarily Shaped Image Segments based on a
generalized Orthogonal Transform", Gilge et.al, Signal Processing: Image
Communication vol. 1 1989, etc..
In "Method of Picture Signal Coding" of Japanese Application Number
6-271542, a procedure which can efficiently coded by inserting a pixel value
even for shapes other than rectangle is proposed.
However, in a picture signal coding apparatus mentioned above,
optimizing of quantization of orthogonal transform for a block other than
rectangular shape having the same size is not investigated and there is a
room to improve from a view point of coding efficiency.
In Gilge's procedure, a complex calculation is necessary for deriving a
basic function of an orthogonal transform and it is difficult to realize. In a
coding procedure inserting a pixel value with a simple calculation, however,
there is a room to realize improving a picture quality more by improving an
inserting method.
DISCLOSURE OF THE INVENTION
The present invention aims to present a picture signal coding
apparatus, a picture signal decoding apparatus, a picture signal coding
method and a picture signal decoding method, in which quantization means
to quantize after the blocks having different sizes are orthogonally
transformed is optimized from a view point of bit rate.
A first invention relates to a picture signal coding apparatus, being
supplied with an input picture signal and an effective pixel indication signal
indicating a partial picture to be coded in the input picture signal and for
coding the partial picture of a digital picture signal at every block, and is
characterized by including:
orthogonal transform base generation means for generating an
orthogonal transform base of the input picture signal by the effective pixel
indication signal; orthogonal transform means for orthogonally transforming the input
picture signal by the orthogonal transform base generated at the orthogonal
transform base generation means; DC normalization coefficient calculation means for deriving a DC
normalization coefficient for normalizing a DC component of a transformed
output signal of the orthogonal transform means from the effective pixel
indication signal; and quantization means for multiplying a quantization step by the DC
normalization coefficient obtained at the DC normalization coefficient
calculation means for a DC component in an orthogonal transform signal
supplied from the orthogonal transform means and quantizing the
orthogonal transform signal by a standard quantization step for a non DC
component in the orthogonal transform signal supplied from the orthogonal
transform means.
A second invention relates to a picture signal decoding apparatus,
being supplied with a coded signal by a picture signal coding apparatus
controlling a quantization step at quantizing a DC component which is a
result of an input picture signal orthogonally transformed by an effective
pixel indication signal indicating a partial picture to be coded and the
effective pixel indication signal and for decoding the coded signal, and is
characterized by including:
decoding means for decoding the coded signal and outputting a decoded
signal; DC normalization coefficient calculation means for deriving a DC
normalization coefficient for normalizing a DC component in the decoded
signal from the decoding means out of the effective pixel indication signal; dequantization means for dequantizing the decoded signal by a
quantization step multiplied by a DC normalization coefficient obtained at
the DC normalization coefficient calculation means for a DC component in
the decoded signal obtained at the decoding means ,and dequantizing the
decoded signal by a standard quantization step for a non DC component in
the decoded signal obtained at the decoding means; orthogonal transform base generation means for generating an
orthogonal transform base of an output signal of the dequantization means
by the effective pixel indication signal; and orthogonal transform means for outputting a picture decoding signal
after orthogonally transforming an output signal of the dequantization
means by the orthogonal transform base.
A third invention relates to a picture signal coding apparatus, being
supplied with an input picture signal, an effective pixel indication signal
indicating a partial picture to be coded in the input picture signal and a DC
component of a reference block and for coding the partial picture in a digital
picture signal at every block and differential coding a DC block at every
block, and is characterized by including:
orthogonal transform base generation means for generating an
orthogonal transform base of the input picture signal by the effective pixel
indication signal; orthogonal transform means for orthogonally transforming the input
picture signal by the orthogonal transform base; DC normalization coefficient calculation means for deriving a DC
normalization coefficient for normalizing a DC component of a transformed
output signal of the orthogonal transform means out of the effective pixel
indication signal; DC prediction coding means for multiplying the DC component of a
reference block by the DC normalization coefficient calculated at the DC
normalization coefficient calculation means and outputting a DC differential
component which is a difference between the obtained DC component and
the DC component of the reference block; and quantization means for quantizing a DC differential component in the
output of the DC prediction coding means and an AC component in the
output signal of the orthogonal transform means.
A fourth invention relates to a picture signal decoding apparatus,
being supplied with a coded signal coded at a picture signal coding
apparatus which a DC component orthogonally transformed from an input
picture signal compensates a DC component of a reference block by using an
effective pixel indication signal indicating a partial picture to be coded,
the effective pixel indication signal and a DC component of the reference
block and for decoding the coded signal, and is characterized by including:
decoding means for decoding the coded signal and outputting the
decoded signal; dequantization means for dequantizing the decoded signal of the
decoding means; orthogonal transform base generation means for generating an
orthogonal transform base of the output signal of the dequantization means
by the effective pixel indication signal; DC normalization coefficient calculation means for deriving a DC
normalization coefficient for normalizing a DC component in the decoded
signal output from the decoding means out of the effective pixel indication
signal; DC prediction decoding means for multiplying a DC component of a
reference block by the DC normalization coefficient, adding it to a DC
differential component of the output of the dequantization means and
making the sum a DC component; and orthogonal transform means for orthogonally transforming the DC
component output from the DC decoding means by the orthogonal transform
base and an AC component output from the dequantization means and
outputting a picture decoded signal.
A fifth invention is characterized by that the DC normalization
coefficient calculation means in the first and third inventions makes the root
of the ratio of the number of pixels of a partial picture to be coded to the total
number of pixels of the block a DC normalization coefficient.
A sixth invention is characterized by that the DC normalization
coefficient calculation means in the second and fourth inventions makes the
root of the ratio of the number of pixels of a partial picture to be coded to the
total number of picture elements of the block a DC normalization coefficient.
A seventh invention relates to a picture signal coding apparatus, being
supplied with an input picture signal and an effective pixel indication signal
indicating a partial picture to be coded of the input picture signal and for
coding a partial picture of a digital picture signal at every block, and is
characterized by including:
orthogonal transform base generation means for generating an
orthogonal transform base of the input picture signal by the effective pixel
indication signal; orthogonal transform means for orthogonally transforming the input
picture signal by the orthogonal transform base generated at the orthogonal
transform base generation means; AC normalization coefficient calculation means for deriving an AC
normalization coefficient for normalizing an AC component of the
transformed output signal of the orthogonal transform means from the
effective pixel indication signal; and quantization means for multiplying a quantization step by the AC
normalization coefficient and quantizing the output signal of the orthogonal
transform means.
An eighth invention is characterized by that the AC normalization
coefficient calculation means makes an inverse of the number of pixels of a
partial picture to be coded an AC normalization coefficient.
A ninth invention relates to a picture signal decoding apparatus, being
supplied with a signal coded at a picture signal coding apparatus controlling
a quantization step of an AC component of an input picture signal
orthogonally transformed by an effective pixel indication signal indicating a
partial picture to be coded and the effective pixel indication signal and for
decoding the coded signal and is characterized by including:
decoding means for decoding the coded signal and outputting a decoded
signal; AC normalization coefficient calculation means for deriving an AC
normalization coefficient for normalizing an AC component in the decoded
signal output from the decoding means out of the effective pixel indication
signal; dequantization means for multiplying the quantization step of an AC
component in the decoded signal obtained at the decoding means by an AC
normalization coefficient derived at the AC normalization coefficient
calculation means and dequantizing the decoded signal; orthogonal transform base generation means for generating an
orthogonal transform base of the output signal at the dequantization means
by the effective pixel indication signal; and orthogonal transform means for orthogonally transforming the output
signal of the dequantization means by the orthogonal transform base and
outputting a picture decoded signal.
A tenth invention is characterized by that the AC normalization
coefficient calculation means makes an inverse of the number of pixels of a
partial picture to be coded an AC normalization coefficient.
An eleventh invention relates to a picture signal coding apparatus,
being supplied with an input picture signal and an effective pixel indication
signal indicating a partial picture to be coded in the input picture signal and
for coding a partial picture of a digital picture signal at every block and is
characterized by including:
a plurality of pixel value generation means for outputting a
substitution of a pixel value generated by at least two kinds of specified rules
into a pixel value of a partial picture which is unnecessary for coding the
input picture signal by the effective pixel indication signal; a plurality of orthogonal transform means for orthogonally
transforming each output signal of the plurality of pixel value generation
means; selection means for comparing each output of the plurality of
orthogonal transform means and selecting an output which has less coding
quantity; and quantization means for quantizing the output signal of the selection
means.
A twelfth invention is characterized by that the selection means selects
an output signal having less high frequency components after orthogonal
transformation.
A thirteenth invention is characterized by that the selection means
selects an output signal having smaller sum of the absolute values of the
components after orthogonal transformation.
A fourteenth invention relates to a pictures signal coding apparatus,
being supplied with an input picture signal and a transparency signal
expressing a synthesis ratio at synthesizing the input picture signal and the
other picture signals and for coding a digital picture signal at every block
and is characterized by including:
orthogonal transform means for orthogonally transforming the input
picture signal; normalization coefficient calculation means for deriving a
normalization coefficient normalizing a transformed output signal of the
orthogonal transform means out of the transparency signal; and quantization means for multiplying the quantization step by the
normalization coefficient and quantizing an output signal of the orthogonal
transform means.
A fifteenth invention is characterized by that the normalization
coefficient calculation means has a large transparency and makes a
normalization coefficient large against a block including many pixels having
a small ratio used at picture synthesis.
A sixteenth invention relates to a picture signal decoding apparatus,
being supplied with a coded signal, normalizing an orthogonal transform
output signal of the input picture by a transparency signal expressing a
synthesis ratio at synthesizing an input picture signal and the other picture
signals, and the transparency signal and is characterized by including:
decoding means for decoding the coded signal and outputting the
decoded signal; normalization coefficient calculation means for deriving a
normalization coefficient normalizing each component of the decoded signal
out of the transparency signal; dequantization means for multiplying a quantization step of the
decoded signal by the normalization coefficient derived at the normalization
coefficient calculation means and dequantizing the decoded signal; and orthogonal transform means for orthogonally transforming the output
signal of the dequantization means and outputting the picture decoding
signal.
A seventeenth invention is characterized by that the normalization
coefficient calculation means has a large transparency and makes a
normalization coefficient large against a block including many pixels having
a small ratio used at picture synthesis.
A picture signal coding apparatus of the first and third inventions is
what calculates a dynamic range of a DC component after orthogonal
transformation at the DC normalization coefficient calculation means and
eliminates a variation of the DC component by a different orthogonal
transform base. Considering coding of k pieces of pixels at every block,
when N pieces of pixels in an effective pixel area w inside a block having k
pieces of pixels is orthogonally transformed, the energy magnitude of the DC
component is different between area w and the whole block. If only the pixels
in area w are orthogonally transformed, the DC component becomes (N/k)
1/2 times comparing with the case orthogonally transforming the whole block.
Generally, when a picture signal is orthogonally transformed and
coded, a technique to code a difference of the DC components between
adjacent blocks is used. Because in a picture signal, the DC components of
the adjacent blocks have a strong correlation, it is tried to improve a
compression efficiency by utilizing that the difference is zero. However,
utilizing a technique that the number of pixels to be coded is different at
every block, the DC difference does not concentrates near zero because the
DC component varies.
When the number of pixels to be orthogonally transformed is N pieces,
a picture signal coding apparatus can eliminate variation of the DC
component between blocks by multiplying a quantization step by (N/k)1/2. In
this case, a coding efficiency at variable length coding means is improved,
where differential coding of the DC component is also included in the
variable length coding means.
A picture signal coding apparatus of the third invention multiplies the
DC component of adjacent blocks by N1/2 and eliminates a DC component
variation due to the different orthogonal transform bases. Thus, a coding
efficiency at the DC prediction signal coding means is increased.
A picture signal decoding apparatus of the second invention and a
picture signal decoding apparatus of the fourth invention derive a DC
normalization coefficient, utilizing an effective picture element indication
signal, and correctly decode a coded signal at a picture signal coding
apparatus of the first invention and a picture signal coding apparatus of the
third invention.
A picture signal coding apparatus of the seventh invention is what
calculates a dynamic range of an AC component after orthogonal
transformation at AC normalization coefficient calculation means and
eliminates a variation of an AC component due to different orthogonal
transform bases. In the case in which orthogonal transformation is done
after inserting a pixel value or the like, because the inserted pixel value has
correlation with the other pixel values, a deviation occurs in a quantization
error. Therefore, if the deviation of the quantized picture element is
eliminated, it is possible to reduce the energy of an average quantization
error lower than that of the other blocks. Let an average quantization error
of a block (the number of pixels is N) not accompanying pixel insertion is
delta, the error at every block when the number of pixels to be coded has only
N pieces of components (inserting M-N pieces of pixels) is close to (N2/M)· δ,
assuming the quantization error is a white Gaussian noise.
Therefore, because the quantization error per pixel is (N 2/M 2)· δ, it
becomes the same quantization error as that of the other blocks by
multiplying the quantization step by M/N. In a picture signal coding
apparatus, because M/N is a real number larger than 1 (including 1), the
number of coding bits can be reduced by multiplying by an AC normalization
coefficient calculated at the AC normalization coefficient calculation means
and dequantizing by the quantization step.
A picture signal decoding apparatus of the ninth invention derives an
AC normalization coefficient by using an effective pixel indication signal and
correctly decodes a signal coded at a picture signal coding apparatus of the
seventh invention.
A picture signal coding apparatus of the eleventh invention, using an
effective pixel indication signal and utilizing that an inserted pixel value is
independent of a decoding procedure of a picture signal decoding apparatus,
generates the inserted pixel value at a plurality of pixel value generation
means, selects a pixel value inserted at a pixel value generation means
which has less coded number of bits after an orthogonal transformation and
can reduce the coded number of bits by variable length coding. A method
inserting an average value of the pixel values and a method generating pixel
values at LPF can be used for pixel value generation means.
A picture signal coding apparatus of the fourteenth invention utilizes
that for a block including pixel having a large transparency and less
influence at picture synthesis, deterioration is visually not remarkable even
for a coarse quantization step. For a block having a large transparency,
normalization coefficient is made large and as a result, the coded number of
bits can be adaptively reduced by making the quantization step coarse.
A picture signal decoding apparatus of the sixteenth invention derives
a normalization coefficient utilizing a transparency signal and correctly
decodes a signal coded at a picture signal coding apparatus of the fourteenth
invention.
BRIEF EXPLANATION OF THE DRAWINGS
FIG. 1 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with a first exemplary embodiment of
the present invention.
FIG. 2 is a block diagram showing a basic composition of a picture
signal decoding apparatus in accordance with a second exemplary
embodiment of the present invention.
FIG. 3 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with a third exemplary embodiment
of the present invention.
FIG. 4 is a block diagram showing a basic composition of a picture
signal decoding apparatus in accordance with a fourth exemplary
embodiment of the present invention.
FIG. 5 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with a fifth exemplary embodiment of
the present invention.
FIG. 6 is a block diagram showing a basic composition of a picture
signal decoding apparatus in accordance with a sixth exemplary
embodiment of the present invention.
FIG. 7 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with a seventh exemplary
embodiment of the present invention.
FIG. 8 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with an eighth exemplary
embodiment of the present invention.
FIG. 9 is a block diagram showing a basic composition of a picture
signal decoding apparatus in accordance with a ninth exemplary
embodiment of the present invention.
THE PREFERRED EXEMPLARY EMBODIMENT OF THE PRESENT
INVENTION
A picture signal coding apparatus in accordance with a first exemplary
embodiment of the present invention is explained below, referring the
drawings. FIG. 1 is a block diagram showing a basic composition of a picture
signal coding apparatus in accordance with a first exemplary embodiment of
the present invention. In the drawing, an effective pixel indication signal 1
corresponding to a picture signal and a picture signal 2 which is made into a
block are supplied to a picture signal coding apparatus. An effective pixel
detection means 3 detects effective pixels which are picture elements to be
coded by effective pixel indication signal 1. An effective pixel information 4
output from effective pixel detection means 3 is supplied to an orthogonal
transform base generation means 6 and a number of pixels signal 5 is
supplied to a DC normalization coefficient calculation means 10 as a signal
indicating the number of effective pixels of the block.
Orthogonal transform base generation means 6 generates an
orthogonal transform base 7 and supplies it to two dimensional orthogonal
transform means 8. Two dimensional orthogonal transform means 8
orthogonally transforms picture signal 2 by orthogonal transform base 7 and
outputs an orthogonal transform signal 9. DC normalization coefficient
calculation means 10 calculates a DC normalization coefficient 11 from
number of pixels signal 5 and supplies it to weighting calculation means 12.
Weighting calculation means 12 calculates a weighting coefficient 13 at
quantization and outputs it and the weighting coefficient 13 is supplied to
quantization means 15 together with a quantization parameter 14.
Quantization means 15 calculates a quantization step from weighting
coefficient 13 and quantization parameter 14, quantizes an orthogonal
transform signal 9 and outputs a quantization value 16. Variable length
coding means 17 codes quantization value 16 in variable length and outputs
a coded signal 18.
The function of a picture signal coding apparatus in accordance with a
first exemplary embodiment composed like above is explained below. A
position of an effective pixel to be coded is indicated by effective pixel
indication signal 1. Orthogonal transform base generation means 6
generates a specified orthogonal transform base 7 from the effective pixel
position in the corresponding block. Two dimensional orthogonal transform
means 8 orthogonally transforms picture signal 2 by using orthogonal
transform base 7 and outputs an orthogonal transform signal 9.
On the other hand, DC normalization coefficient calculation means 10
calculates a DC normalization coefficient 11 from number of pixels signal 5.
For example, the DC normalization coefficient 11 is a function of the
effective number of pixels and the effective number of pixels. Weighting
calculation means 12 calculates a weighting coefficient 13 of the DC
component corresponding to the value of DC normalization coefficient 11.
Quantization means 15 calculates a quantization step from weighting
coefficient 13 and a quantization parameter 14, quantizes orthogonal
transform signal 9 and outputs quantization value 16. Quantization
parameter 14 is a parameter to vary a compression ratio at every block used
for varying the compression ratio by MPEG and if the quantization
parameter is large, coding distortion becomes large but the coded number of
bits becomes small.
As an example of a quantization step, a product of a quantization
parameter 14 and a weighting coefficient 13 is thought. A quantization value
16 generated at quantization means 15 is coded with a variable length at a
variable length coding means 17 and is output as a coded signal 18. At
variable length coding means 17, it is possible not only to code at every block,
but to code a difference from the block input in past with a built-in delay
buffer.
The above signal processing is explained below, using equations. A DC
normalization coefficient calculation means 10 shown in FIG. 1 is provided
for calculating a dynamic range of the DC component after orthogonal
transformation and eliminating a variation of the DC component due to a
different orthogonal transform base. If the number of pixels of a block B is k
pieces and let consider to code each pixel value (Xi; i=1, 2, ...k) at every block.
If an effective pixel area inside block B having k pieces of picture elements is
w and N pieces of pixels existing in effective pixel area w is orthogonally
transformed, the DC component DCw of effective pixel area w is expressed
by Eq.(1).
DCW = 1√N W Xi
If it is orthogonally transformed in whole block B, the DC component,
DCBLOCK is expressed by Eq. (2).
DCBLOCK = 1k BLOCK Xi
Here, if an average value of pixel value Xi is expressed by XAv, each
DC component is expressed by Eq. (3).
DCW = 1√N W Xi = 1√N W XAV = √N XAV DCBLOCK = 1√k BLOCK Xi = 1√k BLOCK XAV = √k XAV
Therefore, only the picture element of effective pixel area w is
orthogonally transformed, the DC component is increased by a ratio
expressed by Eq. (4), compared with the case whole block B is orthogonally
transformed.
Nk
Generally, when a picture signal is orthogonally transformed and
coded, a procedure to code a difference between the DC components of the
adjacent blocks is used. Because the DC components of the adjacent blocks
have a strong correlation, the compression ratio is tried to be improved by
utilizing that the difference between them becomes zero. However, if a
procedure in which the number of pixels to be coded is different from block to
block is used like the above, the difference of the DC components does not
become zero.
In a picture signal coding apparatus in accordance with the first
exemplary embodiment, a variation of the DC component between blocks is
eliminated by increasing (multiplying) the quantization step by the value
expressed by Eq. (5).
Nk
Thus, the coding efficiency at variable length coding means 17
increases. At variable length coding means 17, coding of the difference
between the DC components is also assumed to be included.
Thus, according to the exemplary embodiment, a suitable weighting of
the DC component is made to the corresponding block by effective pixel
indication signal 1, an optimum quantization is done at quantization means
15 and an efficient picture signal coding apparatus can be realized which has
less coded number of bits and is independent on the number of effective
pixels.
A picture signal decoding apparatus in accordance with a second
exemplary embodiment of the present invention is explained below,
referring to FIG. 2. FIG. 2 is a block diagram of a basic composition of a
picture signal decoding apparatus in accordance with the second exemplary
embodiment and the blocks having similar functions to those of the first
exemplary embodiment are numbered with the same reference numbers and
detailed explanation is omitted. A picture signal decoding apparatus of the
second exemplary embodiment derives a DC normalization coefficient by
utilizing an effective pixel indication signal and correctly decodes a coded
signal 18 coded at the picture signal coding apparatus shown in FIG. 1.
In FIG. 2, an effective pixel indication signal 1 is a signal indicating an
effective pixel corresponding to a pixel signal. A coded signal 18 is a signal
coded at a picture signal coding apparatus. An effective pixel detection
means 3 detects an effective pixel which is a pixel to be coded out of effective
pixel indication signal 1 and outputs effective pixel information 4 and a
number of pixels signal 5 expressing the number of pixels of the
corresponding block. Orthogonal transform base generation means 6
generates an orthogonal transform base 7.
DC normalization coefficient calculation means 10 calculates a DC
normalization coefficient 11 from a number of pixels signal 5 and gives it to
weighting calculation means 12. Weighting calculation means 12
calculates a weighting coefficient 13 at quantization. Variable length
decoding means 20 decodes a coded signal 18 and outputs a decoded signal
21. Dequantization means 22 calculates a quantization step from a
quantization parameter 14 and weighting coefficient 13 and outputs a
dequantization value 23. Two dimensional orthogonal transform means 24
orthogonally transforms dequantization value 23 by orthogonal transform
base 7 and outputs a picture decoded signal 25.
The function of a picture signal decoding apparatus of second
exemplary embodiment composed like the above is explained below. The
blocks and the signals numbered with 1-7 and 10-14 shown in FIG. 2 are
similar to those of the first exemplary embodiment and their explanations
are omitted. Coded signal 18 is decoded at variable length decoding means
20, that is reverse of coding at variable length coding means 17 shown in
FIG. 1 and is transformed into a decoded signal 21. Dequantization means
22 calculates a quantization step from weighting coefficient 13 and
quantization parameter 14 for decoded signal 21 and outputs a
dequantization value 23 by dequantizing. The quantization step is similar to
that at quantization means 15 of the first exemplary embodiment.
Two dimensional orthogonal transform means 24 orthogonally
transforms dequantization value 23 and outputs a picture decoded signal 25.
The orthogonal transformation is a reverse transformation of the
transformation at two dimensional orthogonal transformation means 8 of
the first exemplary embodiment.
Thus, according to the second exemplary embodiment, a suitable
weighting of the DC component is done for a corresponding block by using an
effective pixel indication signal 1 and dequantization is done with the same
quantization step as that of the picture signal coding apparatus at
dequantization means 22. Accordingly, a coded signal 18 generated at the
picture signal coding apparatus of the first exemplary embodiment can be
correctly decoded.
A picture signal coding apparatus in accordance with a third exemplary
embodiment of the present invention is explained below, referring to the
drawing. FIG. 3 is a block diagram of a basic composition of a picture signal
coding apparatus in accordance with the third exemplary embodiment. In
FIG. 3, blocks and signals numbered 1-18 are similar to those of the first
exemplary embodiment and their explanations are omitted. A picture signal
coding apparatus of the third exemplary embodiment increases a coding
efficiency of the DC prediction coding means by multiplying the DC
component of the adjacent block by the value expressed by Eq. (5) and
eliminating a variation of the DC component due to a different orthogonal
transform base.
Multiplication means 31 multiplies a DC component 30 of a reference
block by a weighting coefficient 13 and outputs a product signal 32. DC
predicting signal coding means 33 is supplied with product signal 32 and
orthogonal transform signal 9 output from two dimensional orthogonal
transform means 8, calculates a DC differential signal 34 and outputs it.
Synthesis means 35 replaces a DC component of orthogonal transform signal
9 by DC differential signal 34 and outputs an orthogonal transform signal
36.
Only the functions of the different parts of the third exemplary
embodiment composed like the above from those of a picture signal coding
apparatus of the first exemplary embodiment is explained below.
Multiplication means 31 multiplies DC component 30 of the reference block
by weighting coefficient 13. Because weighting coefficient 13 expresses a
variation component of the DC block of the corresponding block due to the
effective number of pixels, product signal 32 is a signal which the variation
component due to the effective number of pixels is normalized and the
prediction efficiency of DC predicting signal coding means 33 becomes large
regardless of the effective number of pixels.
Synthesis means 35 replaces only the DC component of orthogonal
transform signal 9 by the output of predicting signal coding means 33
corresponding to the DC component normalized at the effective number of
pixels. Variable length coding means 17 can increase a coding efficiency by
variable length coding which is general-purpose and independent on the
effective number of pixels.
Thus, according to the third exemplary embodiment, an efficient
picture signal coding apparatus which is regardless of the effective number
of pixels can be realized by normalizing the DC component of a reference
block by weighting coefficient 13 at multiplication means 31 and coding a DC
predicting signal, like the first exemplary embodiment.
A picture signal decoding apparatus in accordance with a fourth
exemplary embodiment of the present invention is explained below,
referring to FIG. 4. FIG. 4 is a block diagram of a basic composition of a
picture signal decoding apparatus in accordance with the fourth exemplary
embodiment and the blocks having similar functions to those of the third
exemplary embodiment are numbered with the same reference numbers and
detailed explanations are omitted. A picture signal decoding apparatus of
the fourth exemplary embodiment decodes a coded signal 18 coded at a
picture signal coding apparatus shown in FIG. 3.
In FIG. 4, the names of blocks and signals numbered with 1-21 and
24-25 are the same as those of the second exemplary embodiment. DC
predicting signal decoding means 40 is supplied with a product signal 32 of a
multiplication means 31, decodes a DC component of a dequantization value
23 and outputs a DC decoding signal 41. Synthesis means 42 outputs a
synthesized signal 43 made by synthesizing a DC component and an AC
component.
The function of a picture signal decoding apparatus of the fourth
exemplary embodiment composed like the above is explained below. The
steps to output a product signal 32 and an orthogonal transform base 7 from
an effective pixel indication signal 1 are the same as the case of a picture
signal decoding apparatus of the second exemplary embodiment. Coded
signal 18 decodes at variable length decoding means 20, in the reverse of
coding of variable length coding means 17 shown in FIG. 3 and outputs a
decoded signal 21.
Dequantization means 22 calculates a quantization step by using a
quantization parameter 14 for a decoded signal 21 and outputs a
dequantization value 23 by dequantization. Because the DC component of
dequantization value 23 is prediction coded, DC prediction decoding means
40 adds a value multiplying the DC component of a reference block by
product signal 32 to the DC component of dequantization value 23 and
outputs a DC coded signal 41.
The DC component of the reference block is assumed to be stored in DC
prediction decoding means 40 for making the explanation of the function
simple and the DC component of the corresponding block is assumed to be
used for decoding the DC component of the following block. Synthesis means
42 replaces the DC component of dequantization value 23 to the DC decoded
signal 41 and outputs a synthesized signal 43. Two dimensional orthogonal
transform means 24 orthogonally transforms synthesized signal 43 and
outputs a picture decoded signal 25.
Thus, according to the fourth exemplary embodiment, coded signal 18
which is made by DC prediction coding can be correctly decoded by
normalizing DC component 30 of the reference block by weighting coefficient
13 at multiplication means 31.
A picture signal coding apparatus in accordance with a fifth exemplary
embodiment of the present invention is explained below, referring to the
drawing. FIG. 5 is a block diagram of a basic composition of a picture signal
coding apparatus in accordance with the fifth exemplary embodiment. In
FIG. 5, blocks and signals numbered 1-18 are similar to those of the first
exemplary embodiment and their explanations are omitted.
An AC normalization coefficient calculation means 50 calculates an AC
normalization coefficient 51 from a number of pixels signal 5 outputted from
an effective pixel detection means 3. A weighting calculation means 52
calculates a weighting coefficient 53 at quantization.
A picture signal coding apparatus in accordance with the fifth
exemplary embodiment calculates a dynamic range of an AC component
after orthogonal transformation at AC normalization coefficient calculation
means 50 and eliminates a variation of the AC component due to a different
orthogonal transform base. In the case in which orthogonal transformation
is done by inserting a picture element value, a deviation occurs in a
quantization error, because the inserted pixel value has a correlation with
the other pixel value. Therefore, if the deviation of the quantized pixel is
removed, it is possible to reduce energy of an average quantization error
more than the other block.
If a deviation of the quantization error of the block (the number of
pixels is N) which does not accompany pixel value insertion is delta, in the
case in which the number of pixels to be coded has only N pieces of
components, that is in the case in which M-N pieces of pixels are inserted,
the error at every block becomes close to a value expressed by Eq. (6),
assuming the quantization error is a white Gaussian noise.
N 2 M δ
Therefore, a quantization error per pixel is expressed by Eq. (7).
N 2 M 2 δ
The quantization error becomes the same as that of the other block by
multiplying a quantization step by the following Eq. (8).
MN
Because the value of Eq. (8) is a real number equal to and more than 1,
the coded number of bits can be reduced by quantizing by a quantization
step multiplied by an AC normalization coefficient calculated at AC
normalization coefficient calculation means 50.
Thus, AC normalization coefficient calculation means 50 calculates an
AC normalization coefficient 51 orthogonally transformed from number of
pixels signal 5. The following function can be realized similarly even if the
DC component is included in AC normalization coefficient 51. For example, a
reverse value of the effective number of pixels of a block is thought as AC
normalization coefficient 51. Weighting calculation means 52 calculates a
weighting coefficient 53 of each coefficient component corresponding to the
value of AC normalization coefficient 51.
Quantization means 15 calculates a quantization step from weighting
coefficient 53 and a quantization parameter 14 and outputs a quantization
value 16. Because the quantization step has an effect to average the
deviations of quantization errors depending on the effective number of pixels,
the coding efficiency can be improved by suitably sharing the number of bits.
As an example of a quantization step, a product of a quantization parameter
14 and a weighting coefficient 53 is thought.
Thus, according to the fifth exemplary embodiment, the AC component
of the corresponding block is suitably weighted by using an effective pixel
indication signal 1, an optimum quantization is done at quantization means
15 and a picture signal coding apparatus can be realized which has less
coded number of bits, is independent on the number of effective pixels and is
efficient.
A picture signal decoding apparatus in accordance with a sixth
exemplary embodiment of the present invention is explained below,
referring to FIG. 6. FIG. 6 is a block diagram of a basic composition of a
picture signal decoding apparatus in accordance with the sixth exemplary
embodiment and the blocks having similar signals and similar functions to
those of the second exemplary embodiment are numbered with the same
reference numbers and detailed explanations are omitted. A picture signal
decoding apparatus of the sixth exemplary embodiment decodes a coded
signal 18 coded at a picture signal coding apparatus shown in FIG. 5.
Only the functions of the different parts of the sixth exemplary
embodiment composed like the above from those of a picture signal coding
apparatus of the second exemplary embodiment is explained below. A
dequantization means 22 is supplied with a decoded signal 21, calculates a
quantization step from a weighting coefficient 53 and a quantization
parameter 14 and outputs a dequantization value 23 by dequantization.
Weighting coefficient 53 is the same as weighting coefficient 53 in the fifth
exemplary embodiment. Therefore, the quantization step is the same as
quantization step 15 in the fifth exemplary embodiment. Although the
picture signal decoding apparatus gives weighting for a DC component in the
second exemplary embodiment, in the sixth exemplary embodiment, it is
different in giving weighting for all coded orthogonal transform components
Thus, according to the sixth exemplary embodiment, the AC component
of the corresponding block is suitably weighted by using an effective pixel
indication signal 1, dequantization is done by the same quantization step as
that of the picture signal coding apparatus and a coded signal 18 of the fifth
exemplary embodiment can be correctly decoded.
A picture signal coding apparatus in accordance with a seventh
exemplary embodiment of the present invention is explained below,
referring to the drawing. FIG. 7 is a block diagram of a basic composition of a
picture signal coding apparatus in accordance with the seventh exemplary
embodiment. In FIG. 7, an effective pixel indication signal 1 corresponding
to a pixel signal and a picture signal 2 made into a block are supplied to a
picture signal coding apparatus. Pixel value generation means 60a and 60b
generate picture signals 61a and 61b, each inserted with a specified value
into a pixel value other than the effective pixel and outputs them.
Two dimensional orthogonal transform means 8a orthogonally
transforms picture signal 61a and outputs orthogonal transform signal 9a.
Similarly, two dimensional orthogonal transform means 8b orthogonally
transforms picture signal 61b and outputs orthogonal transform signal 9b.
Selection means 62 selects one of orthogonal transform signals 9a and 9b
and outputs an orthogonal transform signal 63. Quantization means 15
calculates a quantization step by using a quantization parameter 14 and
outputs a quantization value 16. Variable length coding means 17
transforms quantization value 16 into a coded signal 18 having a variable
length.
The function of a picture signal coding apparatus of the seventh
exemplary embodiment composed like the above is explained below. In the
pixel position made into a block, the position of an effective pixel to be coded
is expressed by an effective picture indication signal 1. Pixel value
generation means 60a and 60b replace a pixel value other than the effective
pixels in the corresponding block into a pixel value by each specified rule
from picture signal 2 (for example, a method to generate a pixel value so that
the high frequency component is small, a method to generate a pixel value by
using an average value in the block or the like) and outputs picture signals
61a and 61b. Picture signals 61a and 61b are orthogonally transformed into
orthogonal transform signals 9a and 9b at two dimensional orthogonal
transform means 8a and 8b, respectively.
Selection means 62 compares orthogonal transform signals 9a and 9b,
selects either of them by a specified rule (for example, selects one having a
less coded number of bits) and outputs an orthogonal transform signal 63. As
a result, a block inserted with a pixels value other than the effective pixel
position at every block can be suitably selected and the coding efficiency can
be increased.
Because orthogonal transform signals 9a and 9b are nearly equal when
all pixels in a block are effective pixels, either one of two dimensional
orthogonal transform means 8a and 8b may be omitted to calculate.
Quantization means 15 calculates a quantization step from a quantization
parameter 14 and outputs a quantization value 16. The quantization value
16 is coded with a variable length at a variable length coding means 17 and
is output as a coded signal 18.
Thus, according to the seventh exemplary embodiment, by using an
effective pixel indication means and by using that the inserted pixel value is
independent on the decoding procedure of a picture signal decoding
apparatus, the coded number of bits can be decreased by generating the
inserted pixel value at a plurality of pixel value generation means 60a and
60b, selecting the pixel value inserted at the pixel generation means having
less coded number of bits after orthogonal transformation and coding with a
variable length. As pixel value generation means 60a and 60b, for example,
it is possible to use a method to insert an average value of the pixel values or
a method to generate a pixel value at LPF.
A picture signal coding apparatus in accordance with a eighth
exemplary embodiment of the present invention is explained below,
referring to the drawing. FIG. 8 is a block diagram of a basic composition of a
picture signal coding apparatus in accordance with the eighth exemplary
embodiment. In FIG. 8, a transparency signal 70 corresponding to a pixel
signal and a picture signal 2 made into a block are supplied to a picture
signal coding apparatus. Transparency calculation means 71 calculates
transparency 72 at every block. Normalization coefficient calculation means
73 calculates normalization coefficient 74 of the corresponding block from
transparency signal 72.
Weighting calculation means 75 calculates a weighting coefficient 76 at
quantization. Two dimensional orthogonal transform means 8 orthogonally
transforms a picture signal 2 and outputs an orthogonal transform signal 9.
Quantization means 15 calculates a quantization step from a quantization
parameter 14 and weighting coefficient 76 and outputs a quantization value
16. Variable length coding means 17 codes with variable length the
quantization value 16 and outputs a coded signal 18.
The function of a picture signal coding apparatus of the eighth
exemplary embodiment composed like the above is explained below.
Transparency information of each pixel corresponding to a picture signal
made into a block is expressed by a transparency signal 70. Two dimensional
orthogonal transform means 8 orthogonally transforms picture signal 2 into
an orthogonal transform signal 9. On the other hand, transparency
calculation means 71 calculates transparency of the corresponding block (for
example, an average transparency, a minimum transparency, etc. of the
corresponding block) from transparency signal 70 and outputs as a
transparency signal 72.
A large value of transparency signal 72 means that it is difficult to
visually influence because the pixel is transparent. In this case,
normalization coefficient calculation means 73 outputs a normalization
coefficient 74 having a coarse quantization step. Quantization means 15
calculates a quantization step from a weighting coefficient 76 and a
quantization parameter 14 and outputs a quantization value 16. Because
the quantization step has an effect to remove a visual influence which is
dependent on a transparency, the coding effect can be increased preventing
large deterioration of picture quality.
Quantization value 16 is variable length coded at variable length
coding means 17 and is output as a coded signal 18. Variable length coding
means 17 not only codes at every block but has a delay buffer in it and can
code the difference from the block input in the past.
Thus, according to the eighth exemplary embodiment, a suitable
weighting of an orthogonal transform is done from a transparency signal 70
and as a result, an optimum quantization is done at quantization means 15
and an efficient picture signal coding apparatus can be realized which has
less coded number of bits, is dependent on the transparency and eliminates
deterioration of picture quality.
A picture signal decoding apparatus in accordance with a ninth
exemplary embodiment of the present invention is explained below,
referring to FIG. 9. FIG. 9 is a block diagram of a basic composition of a
picture signal decoding apparatus in accordance with the ninth exemplary
embodiment, which decodes a coded signal 18 coded at a picture signal
coding apparatus shown in FIG. 8.
In FIG. 9, the blocks and signals numbered with 70-76 are similar to
those of the eighth exemplary embodiment shown in FIG. 8 and their
explanations are omitted. A variable length decoding means 20 decodes a
coded signal 18 and outputs a decoded signal 21. Dequantization means 22
calculates a quantization step from a quantization parameter 14 and a
weighting coefficient 76 and outputs a dequantization value 23. Two
dimensional orthogonal transform means 24 orthogonally transforms
dequantization value 23 and outputs a picture decoded signal 25.
The function of a picture signal decoding apparatus of the ninth
exemplary embodiment composed like the above is explained below. The
functions of the blocks and the signals numbered with 70-76 are the same as
those of the eighth exemplary embodiment and their explanations are
omitted. At variable length decoding means 20, coded signal 18 is decoded in
reverse to the coding at variable length coding means 17 in the eighth
exemplary embodiment and is converted into a decoded signal 21.
Dequantization means 22 calculates a quantization step, using a
decoded signal 21, a weighting coefficient 76 and a quantization parameter
14 and outputs a dequantization value 23 after dequantization. The
quantization step is similar to that at dequantization means 15 of the eighth
exemplary embodiment. Dequantization value 23 is orthogonally
transformed at two dimensional orthogonal transform means 24 and is
output as a picture decoded signal 25. The orthogonal transformation made
at two dimensional orthogonal transform means 24 is similar to reverse
transformation at orthogonal transform means 8 in the first exemplary
embodiment.
According to the ninth exemplary embodiment, a suitable weighting is
done from a transparency signal 70 to the orthogonal transform coefficient of
the corresponding block and reverse quantization is done at reverse
quantization means 22 by a similar quantization step to that of the picture
signal coding apparatus. Thus, a coded signal 18 of the eighth exemplary
embodiment can correctly be decoded.
Because the composing blocks are almost same in the above exemplary
embodiments, they can be used by combination. Because a transparency
signal which transparency is other than 100% means an effective pixel, an
affective pixel indication signal can be made from a transparency signal.
A system in which a picture signal coding apparatus and a picture
signal decoding apparatus in accordance with the present invention are
connected by wireless or wired communication means can be composed. In
this case, a composition is possible in which data of a coded signal coded at
the picture signal coding apparatus is temporarily stored in a server and is
sent to the picture signal decoding apparatus when it is necessary.
AVAILABILITY IN THE INDUSTRIAL FIELD
As explained above in detail, according to the present invention, a
picture signal coding apparatus and a picture signal decoding apparatus can
be realized which are optimized from a view point of bit rate at quantization
means quantizing after orthogonally transforming a block having a different
size.
In the present invention, an efficient method of pixel value insertion is
also adapted and a picture signal coding apparatus and a picture signal
decoding apparatus having a high coding efficiency can be presented.
Further, in the present invention, a picture signal coding apparatus
and a picture signal decoding apparatus having a higher efficiency can be
realized by optimizing quantization means, using transparency information.
In any invention, when a semitransparent partial digital picture is
inserted into one frame of a digital picture, energy distributions of DC
coefficients and AC coefficients of an inserted picture and an inserting
picture become uniform and a coding efficiency after quantization is
improved.
A system having a good transmission efficiency can be realized by
composing a system connecting a picture signal coding apparatus and a
picture signal decoding apparatus of the present invention by
communication means.
NOTATIONS:
- 1
- effective pixel indication signal
- 2
- picture signal
- 3
- effective pixel detection means
- 4
- effective pixel information
- 5
- number of pixels signal
- 6
- orthogonal transform base generation means
- 7
- orthogonal transform base
- 8, 8a, 8b
- two dimensional orthogonal transform means
- 9, 9a, 9b
- orthogonal transform signal
- 10
- DC normalization coefficient calculation means
- 11
- DC normalization coefficient
- 12
- weighting calculation means
- 13
- weighting coefficient
- 14
- quantization parameter
- 15
- quantization means
- 16
- quantization value
- 17
- variable length coding means
- 18
- coded signal
- 20
- variable length decoding means
- 21
- decoded signal
- 22
- dequantization means
- 23
- dequantization value
- 24
- two dimensional orthogonal transform means
- 25
- picture decoded signal
- 30
- DC component of reference block
- 31
- multiplication means
- 32
- product signal
- 33
- DC prediction coding means
- 34
- DC differential signal
- 35, 42
- synthesizing means
- 36
- orthogonal transform signal
- 40
- DC prediction decoding means
- 41
- DC decoded signal
- 43
- synthesized signal
- 50
- AC normalization coefficient calculation means
- 51
- AC normalization coefficient
- 52
- weighting calculation means
- 53
- weighting coefficient
- 60a, 60b
- pixel value generation means
- 61a, 61b
- picture signal
- 62
- selection means
- 63
- orthogonal transform signal
- 70
- transparency signal
- 71
- transparency calculation means
- 72
- transparency signal
- 73
- normalization coefficient calculation means
- 74
- normalization coefficient
- 75
- weighting calculation means
- 76
- weighting coefficient